Distributed processing applications (e.g., a Hadoop Map-Reduce application) can execute workloads over a cluster of compute nodes, such as a plurality of servers. The compute nodes may be interconnected by a network, such as a software defined network. The compute nodes send messages to each other in order to coordinate processing of the workloads.